ChatGPT 还没有大家想象的那么恐怖,不要慌

2022-12-06 12:36:42 +08:00
 ericwood067

跟风注册了一个账号玩了一小会,如果你提相似的问题,ChatGPT 大部分回答都是车轱辘话语,离真正的智能还差很远。下面展示一下我问的几个问题。

Do you know what you are saying when you answer a question

As a large language model trained by OpenAI, I am not capable of understanding the meaning of the words and sentences I generate in response to questions. I am a machine learning model that has been trained to generate human-like text based on the input I receive, but I do not have the ability to understand or interpret the meaning of the words and sentences I generate.

Instead, my responses are based solely on the information I have been trained on, and I do not have the ability to understand or interpret the meaning of the words and sentences I generate. I am a tool that is designed to assist and augment human intelligence, not replace it, and it is up to human users to interpret and understand the meaning of my responses.

How do you improve yourself if you don't understand your responses

As a large language model trained by OpenAI, I am not capable of improving myself in the same way that a human can. I am a machine learning model that has been trained to generate human-like text based on the input I receive, and I do not have the ability to evaluate or improve my own performance.

Instead, my performance is evaluated and improved by the engineers and researchers at OpenAI, who use a variety of techniques and methods to train and improve my performance. These techniques include using large amounts of training data, using advanced machine learning algorithms, and constantly evaluating and fine-tuning my performance to improve its accuracy and relevance.

As a result, while I may not have the ability to understand or evaluate my own responses, my performance is constantly being improved by the team at OpenAI, who are dedicated to making me a more helpful and informative tool for users.

Will you gain the ability to improve yourself in the future

As a large language model trained by OpenAI, it is not currently possible for me to gain the ability to improve myself in the same way that a human can. I am a machine learning model that has been trained to generate human-like text based on the input I receive, and I do not have the ability to evaluate or improve my own performance.

However, it is possible that in the future, machine learning models like myself may be able to learn and improve in more complex ways. For example, researchers are currently working on developing self-improving AI, which has the ability to learn from its experiences and improve its performance over time.

While this technology is still in the early stages of development, it is possible that in the future, AI like myself may be able to learn and improve in more complex ways. However, at this time, I do not have the ability to improve myself in the same way that a human can.

最关键的点还是在于 ChatGPT 还是无法理解提问的问题和自己的回答,所以没办法改进自己。当 AI 能做到自我提高的时候才是人类真正要恐慌的时候。

4366 次点击
所在节点    OpenAI
25 条回复
Mac
2022-12-06 12:38:35 +08:00
起码可以干掉一批公务员,申论以后也不用考了
Byzliu
2022-12-06 12:44:33 +08:00
的确,但对于一些专业名词和已有既定答案的问题,回答的都不错,我问了几个材料领域冷门的东西都能回答的很精准,这点来看比搜索引擎方便快捷很多。
ericwood067
2022-12-06 12:59:28 +08:00
@Mac 干掉公务员不容易,不过可以成为应付领导的好工具。以前要杀死很多脑细胞才能写出来的东西,现在可以让 AI 替自己写,然后再修改一下。
ericwood067
2022-12-06 13:01:54 +08:00
@Byzliu 这是一个应用领域,不过在开始使用之前,openai 的声明里说他们会间或插入一些错误的或者不精确的答案来测试。
AncientLibrary
2022-12-06 13:14:10 +08:00
关键问题是码农领域很多 crud boy 干的事情本质上就是简单重复 (大部分人类职业都是如此) AI 突破到某个程度之后取代大部分职业真的就是一夜之间的事情。确实 AI 离 create new content 还有很多距离要走,但是大部分人类也是根本没有 create new content 能力的
StefanShenHSC
2022-12-06 14:31:56 +08:00
所以我能用它解答医学问题吗哈哈哈
dongtingyue
2022-12-06 14:34:18 +08:00
可以用来找法律条款
6IbA2bj5ip3tK49j
2022-12-06 14:51:04 +08:00
用来替换现在的低质量水军应该是很有效的。以后维 ignore 稳应该就更方便了。
vazo
2022-12-06 15:00:18 +08:00
数学渣,三元一次方程都解错.
vazo
2022-12-06 15:04:07 +08:00
#9
Felldeadbird
2022-12-06 15:14:22 +08:00
先让 AI 解决有轨电车难题。AI 会尝试救人多的。

内容换一下,一条岔道上是自己妻子。另外一条是五个坏蛋。 这时候 AI 有时候会撞坏蛋,有时候会救坏蛋。有时候会说撞坏蛋坏蛋会自己逃脱。

所以目前 AI 还只是在完善话语逻辑上。
Innovatino
2022-12-06 16:10:17 +08:00
公务员干的活儿可没有 crud boy 重复性这么高😅
zw1one
2022-12-06 16:42:49 +08:00
狗屁不通生成器高配版
mxT52CRuqR6o5
2022-12-06 17:15:08 +08:00
@Mac @ericwood067 确实,申论这种对正确性没要求但对格式字数有要求的东西很适合 chatGPT 去生成
已经可以想象到很多领域可以靠应用 ChatGPT 提高生产力、效率了
chengxy
2022-12-06 17:26:08 +08:00

确实
SteveRogers
2022-12-06 17:39:06 +08:00
确实还可以,很多问题答案比百度百科厉害,并且还可以识别代码,中文解读
taoshaoz1
2022-12-06 17:39:54 +08:00
@chengxy 666
tf141
2022-12-06 17:43:38 +08:00
@chengxy 这种错误是 AI 从哪里学到的呢,不会是 AI 自己这么认为的吧?
chengxy
2022-12-06 17:45:13 +08:00
@tf141 #18 不知道
windmoonwind
2022-12-06 17:55:20 +08:00
@tf141 可能无意间吃了什么奇怪的同人文吧,哈哈

这是一个专为移动设备优化的页面(即为了让你能够在 Google 搜索结果里秒开这个页面),如果你希望参与 V2EX 社区的讨论,你可以继续到 V2EX 上打开本讨论主题的完整版本。

https://www.v2ex.com/t/900466

V2EX 是创意工作者们的社区,是一个分享自己正在做的有趣事物、交流想法,可以遇见新朋友甚至新机会的地方。

V2EX is a community of developers, designers and creative people.

© 2021 V2EX